Re: [PATCH v4] regulator: MAX77686: Add Maxim 77686 regulator driver

[jonghwa3.lee@samsung.com]

On 2012년 05월 30일 21:08, Yadwinder Singh Brar wrote:

> Hi Jonghwa,
> 
> On Wed, May 30, 2012 at 4:07 PM,  <jonghwa3.lee@samsung.com> wrote:
>> Hi Yadwinder,
>>
>> I'm sorry for late reply. I understand the problem you pointed out, but
>> i don't agree with you all.
> 
> Sorry,I think you didn't get my points. Lets forget my code and talk
> about this code now.
> 
>>>>>> +
>>>>>> +       for (i = 0; i < MAX77686_REGULATORS; i++) {
>>>>>> +               if (pdata)
>>>>>> +                       init_data[pdata->regulators[i].id] =
>>>>>> +                                                pdata->regulators[i].initdata;
> 
> In case we have a list of 5 regulators only in pdata, than what will
> happen here when i > 5 ???
> 


You're right, it has bug. How do you think that change the condition to
(pdata && i < pdata->num_regulators)?

>>>>>
>>>>> I  think we can directly use  pdata->regulators[i].initdata instead of
>>>>> init_data[i].
>>>>> In case if pdata is not their we can use same instance of
>>>>> init_data(default)  for all regulators.
>>>>>
>>>>
>>>>
>>>> This if for some situation that pdata's initdata doensn't line up. When
> 
>>>>>> +               config.init_data = init_data[i];
>>>>>> +               rdev[i] = regulator_register(&regulators[i], &config);
> 
> In case pdata->regulators[0] is not the first regulator(i.e id > 0), then
> will it get proper initdata for regulators[0] before registering ????
> 


Yes, because above code replaces pdata->regalator's initdata to proper
position of initdata array referencing regulator's id.

>>>
>>> Ok, but I think this not right place for sorting (sorting is not taking
>>> place.) You have to sort it before entering in loop for registering
>>> regulators.
>>>
>>>> user sets only initdata considered it being used, there may be
>>>> regulators not having initdata, also its order is not clear. So for
>>>
>>> Ok, I think this is a bug in present driver also, because
>>> without checking pdata->num_regulators, you are running in
>>> loop  for (i = 0; i < MAX77686_REGULATORS; i++)
>>> where MAX77686_REGULATORS should be equal to
>>> pdata->num_regulators for this driver to work fine.
>>>
>>
>>
>> I think we have same variable num_regulators but use differently. In my
>> code, it represents number of regulators to be used actually, but in
>> your code it equals to total number of regulators. Since it has
> 
> not exactly.
> 
>> different meaning, it doesn't have to same with MAX77686_REGULATORS.
>> MAX77686_REGULATORS is macro which indicates total number of regulators
>> in max77686, and it equals to ARRAY_SIZE(regulators). Even if they are
>> not same, it's not a bug because we want to register all regulators
>> whether it will be used or not.
>>
>>
>>> If we consider a case pdata->num_regulators is
>>> equal to MAX77686_REGULATORS and initdata is
>>> not their(i.e. NULL) than I think it will initialise
>>> init_data[pdata->regulators[i].id to NULL, which again will be a bug.
>>>
>>>> those state, i think just using temporary array which satisfies
>>>> regulator's id order is fine while it can't use pdata's initdata directly.
>>>>
>>>
>>> If I am not wrong, I think we can also sort pdata's initdata also using
>>> kernel's sort api and use one instance of (default)initdata for
>>> all unused or uninitialized regulators in platform file.
>>>
>>
>>
>> If init_data references to NULL, it will be ignored while
>> register_regulators() does initialize. Thus it doesn't make any problem.
>>
>> I'm afraid of using Kernel's sort API because of its overhead. Do you
> 
> I don't think it's overhead will matter more than that of allocating a
> new array and than
> sorting it here.
> 
>> think it will be better to use them? If you mind that init_data has been
>> dynamic allocated, it can be modified to a static pointer array.
>>
> 
> No, their is no problem with dynamic.
> Anyways, I had just suggested you to use pdata->regulators[i].initdata.
> 


So, to sum up to this, you think it is better to sort pdata->regulators
by its id before entering loop and just use pdata->regulators directly,
right? Okay, I'll do modify it.


>  Regards,
>  Yadwinder.
> --
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at  http://www.tux.org/lkml/
>